Listing up possible scenarios and building rules around these can be one way, but since self driving will continuously learn and advance itself it may be important and interesting to establish the 'right'(what is right? is another question) foundational values into AI directly. Like human, AI would need to have its own value and learn based on it.